specification: FinOps Framework
specificationVersion: '1.0'
schema: https://www.finops.org/framework/
provider: Inferless
providerId: inferless
created: '2026-06-20'
modified: '2026-06-20'
reconciled: true
tags:
- AI
- ML Inference
- Serverless GPU
- Model Deployment
- Inference
- FinOps
- Cost Management
- FOCUS
description: >-
  FinOps view of Inferless spend. Inferless bills per second of GPU compute
  consumed while serving inference, metered by GPU machine type (T4, A10, A100)
  and instance class (shared/fractional vs dedicated). Cost accrues only while a
  request runs; with min_replica set to 0, idle deployments incur no charge.
  Optimization centers on right-sizing the GPU, tuning concurrency and replica
  bounds, and scaling to zero.
notes: >-
  Per-second machine rates are subject to change; verify against the Inferless
  pricing page during reconciliation.
sources:
- https://www.inferless.com/pricing
- https://docs.inferless.com
- https://focus.finops.org/focus-specification/v1-3/
alignedWith:
  framework: FinOps Foundation Framework
  frameworkUrl: https://www.finops.org/framework/
  dataSpec: FOCUS
  dataSpecVersion: '1.3'
  dataSpecUrl: https://focus.finops.org/focus-specification/v1-3/
publisherName: Inferless
serviceCategory: AI and Machine Learning
billingModel:
  pricingCategory: Usage-Based
  billingFrequency: Monthly
  billingCurrency: USD
  chargeCategories:
  - Usage
  - Purchase
  - Adjustment
focusColumns:
  ServiceName: Inferless Serverless GPU Inference
  ServiceCategory: AI and Machine Learning
  ProviderName: Inferless
  PublisherName: Inferless
  InvoiceIssuerName: Inferless
  BillingCurrency: USD
  ChargeCategory: Usage
  PricingCategory: Usage-Based
meters:
- name: gpu_seconds
  description: Seconds of GPU compute consumed serving inference, billed per machine type and instance class.
  unit: seconds
  aggregation: sum
  dimensions:
  - account
  - workspace
  - model
  - machine_type
- name: shared_instance_seconds
  description: GPU seconds on shared (fractional) instances at the lower per-second rate.
  unit: seconds
  aggregation: sum
  dimensions:
  - account
  - model
  - machine_type
- name: dedicated_instance_seconds
  description: GPU seconds on dedicated instances reserving the full GPU.
  unit: seconds
  aggregation: sum
  dimensions:
  - account
  - model
  - machine_type
- name: free_credit
  description: Onboarding free credit (hours) applied against early GPU usage.
  unit: hours
  aggregation: sum
  dimensions:
  - account
principles:
- name: Visibility
  description: Track per-model GPU-second burn and model logs; attribute spend by workspace and deployment.
- name: Allocation
  description: Tag deployments per workload/team via workspaces; map machine type and instance class to internal cost centers.
- name: Optimization
  description: Right-size the GPU (T4 vs A10 vs A100); set min_replica to 0 to scale to zero when idle; tune container_concurrency and max_replica; prefer shared instances for infrequent workloads and dedicated for steady high-throughput.
- name: Accountability
  description: Assign owners per model/workspace; review GPU-second spend monthly against budget and negotiate Enterprise rates at volume.
maintainers:
- FN: Kin Lane
  email: kin@apievangelist.com